Introduction

Food product quality and safety depends on the microbial load and absence of pathogenic bacteria found in production areas [1]. Therefore, cleaning and disinfection processes (C&D) play a key role in the quality control system of food processing plants [2]. Surfaces in the food processing areas that come in contact with products may act as a contamination source of spoilage and pathogenic bacteria [3]. Consequently, these bacteria may reach products during food production directly from conveyor belts, slicers, utensils, or biofilms on surfaces [4, 5].

Biofilms can be an important source of food contamination and lead to economic losses due to spoilage or issues with food safety [6, 7]. Bacteria forming biofilms are protected against environmental stress conditions, such as cleaning, disinfection, drying, high temperature, and low-nutrient conditions, which ensure bacterial viability and persistence in the food production areas [8, 9]. The ability of some food-associated pathogens to attach to food processing surfaces and to form biofilm has been previously reported for pathogens such as Listeria monocytogenes [10] and Staphylococcus aureus [11], as well as for spoilage bacteria such as Pseudomonas spp. and Lactobacillus spp. [12, 13].

Several studies addressing biofilms in the food industry have been focused on specific pathogenic bacteria and have contributed important knowledge about their biology and prevalence [14, 18, 20]. However, the impact of bacterial interactions in food processing environments must be taken into account, since these interactions may promote and improve biofilm formation [15, 16]. Interactions in multispecies biofilms enhance protection of species that are less tolerant to stress factors and have low ability of biofilm formation [17]. To date, the greater ability of some bacterial genera to produce biofilms, compared with others, has been well documented. For instance, some strains of Pseudomonas spp. have been shown to enhance biofilm formation by L. monocytogenes [18]. Likewise, Oliver et al. (2010) indicated that Escherichia coli O157:H7 strain was protected by a biofilm formed by an Acinetobacter calcoaceticus strain isolated from meat processing facilities [19]. Furthermore, both Pseudomonas and Acinetobacter genera are bacteria that can be frequently found on food processing surfaces [9, 20].

Traditionally, the study of microorganisms has been concentrated on analyzing single species in pure culture [12]. However, a significant number of microorganisms cannot be isolated and studied in the laboratory with current conventional microbiological methods. Microbiological culture generally favors the recovery of organisms adapted to laboratory conditions; however, these may not necessarily to be the dominant or the most influential organisms in the environment [14]. Therefore, in order to better understand microbial communities, there is a need for traditional techniques (microbiological culture) to be supplemented with other culture-independent strategies such as metagenomics, which allows the identification of microorganisms recovered directly from their own environment. Using different DNA markers such as 16S, 18S, and ITS (internal transcribed spacer), the metagenomic approach allows the simultaneous study of many organisms without the need of individually isolating each one of them [21].

Identifying bacteria that form biofilms is important not only to determine the resident “Bacteriota” and to understand microbial communities but also to evaluate sanitation processes and microbiological sampling in food processing industries [22, 23]. Therefore, the aim of this study was to characterize, using 16S rRNA sequencing, the bacterial communities from biofilms found in four food processing plants.

Materials and methods

Food processing plants

Four different food processing plants located in two of the departments of Colombia were evaluated. The plants were identified as P1, P2, P3, and P4. P1 and P2 processed ready-to-eat meal products such as lasagna, ravioli, spaghettis, and pizzas. On the other hand, P3 and P4 processed meat products such as mortadella, ham, meatballs, and sausage. The average temperature and humidity of the plants were 12.5 °C and 67%, respectively. A quaternary ammonium-based product was used for the disinfection process in all plants.

Sampling

One hour after C&D processes, a total of 50 samples were collected from packing and production areas. The samples on each plant were taken the same day, but on different days among plants. The sampled areas were in contact with the food produced in the four plants described above. Sampled surface materials were stainless steel 39 (78%), rubber 6 (12%), aluminum 2 (4%), sailcloth 2 (4%), and iron 1 (2%). Biofilms were tested according to two criteria; the first criterion was the observation of changes in the appearance of surfaces from mesons or machines. Those changes were the presence of pale yellow or colorless aggregates (flocks or granules) and plates (opaque) [24]. The second criterion was the confirmation of visually identified biofilms using BioFinder reagent (itram® Hygiene, Spain), following manufacturer’s instructions. Sampling was carried out using sterile FloQ Swabs (COPAN, Italy). Every sample was taken by scraping and brought to 1 ml of sterile water nuclease-free (Roche, Germany), which were transported at 4 °C at the laboratory and stored at − 20 °C until processing (Table 1).

Table 1 Number of points sampled by plant
Full size table

DNA extraction

The collected bacteria were harvested by centrifugation at 14,000×g for 10 min at 4 °C. DNA extraction was performed using High Pure Template kit (Roche, Germany), following manufacturer’s instructions. Finally, samples were frozen at − 80 °C until analysis.

Amplicon-based metagenomic profiling of bacterial communities

Bacterial identification was performed using next-generation sequencing (NGS) using the Illumina Miseq platform and targeting V4 hypervariable region of the 16S rRNA gene. DNA sequencing was outsourced to Research Testing Laboratory (RTL. Lubbock, TX, USA). The following primers were used: 515F-806R as described in a previous study [25]. Three steps were followed after result analysis; denoising to correct error reads using USEARCH [26]. Then, the reads were organized according to size in order to eliminate duplicated sequences. Chimeric sequences were eliminated using the UCHIME software [27]. Operational taxonomic unit (OTU) selection was performed using the UPARSE OTU selection algorithm (identity threshold 97%) [28] to classify the large number of clusters into OTUs. In order to determine the taxonomic information, Greengenes database was used.

Analysis of the bacterial composition

Microbial community composition comparisons were based on pairwise dissimilarity measurements. A variance-stabilizing transformation for relative abundance matrix was applied for dispersion reduction; values were showed in arcsin√x [29].

To assess difference in bacterial community composition, richness and abundance were studied according to alpha diversity by using Shannon index and β-diversity based on SIMPER analysis in order to know the bacteria that contributed more to microbiota composition between plants. We also investigated indicator bacteria within plants through IndicSpecies. Hierarchical clustering was performed on the transformed dataset based on the Bray-Curtis (B-C). These analyses were performed using the R software version 3.5.1.

Venn diagram for description OTUs at family and gender level in each of the four plants was performed using Venny 2.1 program (http://bioinfogp.cnb.csic.es/tools/venny/).

Results

Taxonomic profiling of bacterial communities

A total of 1,426,157 16S rRNA sequences were generated by NGS from the samples collected at the four processing plants studied. Of these, 87.7% (1,250,157) were clustered into some OTUs. The remaining 12.3% (175,585) had no hits associated with OTUs. A total of four samples did not yield amplified product (two samples in P1 and two samples in P3), probably due to DNA degradation, presence of an inhibitor substance in the sample, or that there was no bacterial DNA.

Four phyla were dominant and constituted more than 94% of the identified phyla across all four processing plants: Proteobacteria, Firmicutes, Actinobacteria, and Bacteroides. The most frequent phylum at P1, P3, and P4 was Proteobacteria, representing 47.3%, 51.7%, and 45.3% of OTUs, respectively. On the other hand, in P2, the dominating phylum was Firmicutes with 50.3% of OTUs (Table 2).

Table 2 Main OTUs at phylum level found across the four plants
Full size table

Identification of exclusive and core microbiota across food processing plants

Our analysis identified a total 102 families and 189 genera. Of the 102 families identified, 37 (36.3%) were common across all plants, and 22 (21.6%) were specific. P3 had the highest number of unique families, nine (8.8%), followed by P1, P2, and P4 with six (6.9%), four (3.9%), and two (2%), respectively (Fig. 1a). The most frequent families found were Pseudomonaceae and Moraxellaceae with a 93.5% (43/46) in overall plants, followed by Lactobacillaceae with 89.1% (41/46), and Streptococcaceae family with 78.2% (36/46) (Supplementary S1).

Fig. 1
figure 1

Venn diagrams. The letter and number together outside the ovals and their colors represent each plant. The numbers in intersections indicate common families or genera among plants. The numbers outside of the intersections are the families or genera found only in those plants. a Families identified across plants. b Genera identified across plants

Full size image

Of the 189 OTUs identified at the genus level, 50 (24.5%) genera were shared across all four plants, whereas 70 (37.0%) were specific, being P1 the one with the most unique genera, in total 26 (13.8%), followed by P3 with 24 (12.7%), and P2 and P4 exhibited the same number, 10 (5.3%) (Fig. 1b).

Dominant genera identified in biofilm

Regarding the relative abundance and frequency analysis, Pseudomonas spp. had the highest frequency and relative abundance. This genus was identified in 35.7% of samples from P1, and 50% for P3, and had an average relative abundance of 50.5% and 43.9%, at P1 and P3, respectively. In addition, SIMPER analysis showed that Pseudomonas genus was more common between P1–P2 and P1–P3, with highest similarity percentage, while in P2, Lactobacillus genus was the dominating bacteria with a frequency of 57.1%, and an average relative abundance of 38.6%. Likewise, this genus was the one contributing the most to the similarity between P2–P3. In P4, the highest percentage was for Acinetobacter spp., 66.7%, which had an average relative abundance of 72.9%, contributing to most similarity between P2–P4 and P3–P4 (Fig. 2) (Supplementary S2).

Fig. 2
figure 2

Ten main genera found in each plant according to relative abundance. P1, P2, P3, and P4 represent each food processing plant (X–axis). Color boxes represent bacterial abundance (Y–axis)

Full size image

Resident lactic acid bacteria across food processing plants:

Lactic acid bacteria (LAB) were present in all of the analyzed food processing plants. For example, Streptococcaceae family was the most prevalent in P1 and was represented by Lactococcus spp. and Streptococcus spp., both with the same frequency (78.6%). Lactobacillaceae was the second family most frequently found at P1, being Lactobacillus spp. (71.4%) and Leuconostoc spp. (64.3%) the genera most frequently identified; however, Leuconostoc was recognized as the indicator species at P1 (P < 0.05) (Supplementary S3).

Likewise, two families were dominant in P2, Lactobacillaceae and Streptococcaceae. In this plant, Lactobacillus and Streptococcus genera were the most frequently identified (78.6%). However, Lactobacillus was the indicator species for P2 (P < 0.05) (Supplementary S3). On the other hand, the Streptococcaceae family with the Streptococcus genus (83.3%) and the Alicyclobacillaceae family with Alicyclobacillus genus (66.6%) were the predominating ones in P3. Unlike the other plants, P4 was where less LAB were identified, being Alicyclobacillus genus the most frequent (33.3%).

Enterobacteriaceae across plant:

Enterobacteriaceae family members were also identified. Escherichia spp., Enterobacter spp., Cronobacter spp., Hafnia spp., Morganella spp., and Serratia spp. were some of the genera found in this study. Escherichia spp. were present in all plants, being the most frequently found in P1 (64.2%), followed by P2 (57.1%), P3 (50%), and P4 (25%).

According to our results, pathogenic bacteria associated with foodborne diseases, such as L. moncytogenes, Salmonella spp., S. aureus, Campylobacter spp., or Bacillus subtilis, were not identified in our samples.

Diversity analysis

Analysis of alpha diversity at the family level according to the Shannon index showed that the most diverse plant was P3 with an average of 1.82, followed by P1 with 1.59, P2 with 1.46, and P4 with 1.20 (Supplementary S4). Moreover, P3 was the plant with the greatest number of indicator species, among these, Sphingomona spp., which exhibited the highest strength (90.8%) and statistical significance (P < 0.05).

Beta diversity under Bray-Curtis analysis showed apparently two large groups. P1 and P2 were clustered within group one, whereas the P3 and P4 were clustered within group two (Fig. 3).

Fig. 3
figure 3

Communities at families level by Bray-Curtis analysis. Red square indicates the families clustered into two groups. The first number corresponds to plant (1-X); the second number indicates the point code sampled (X-1)

Full size image

Discussion

The implementation of next-generation sequencing methodologies has improved the study of microbial diversity on different environments [30]. These tools have complemented conventional microbiological techniques [31], which are insufficient to study the biodiversity of microorganisms, since most bacteria cannot be isolated and identified by microbiological culture [32]. In addition, evidence has shown that the absence or inappropriate use of neutralizing substances, as well as unsuitable growing conditions (incubation temperature and culture media), could lead to an underestimation of the bacterial diversity and abundance [33]. The metagenomic approach implemented in this study allowed to determine both diversity and abundance of bacteria that were part of the analyzed biofilms [34, 35].

It has been previously described that resident microbiota in food plants will depend on environmental conditions (air flow, temperature, and humidity), C&D process, food origin (animal, vegetable), geographical region, and even on human microbiota [36]. Beta diversity analysis at the family level showed that bacterial communities forming biofilms found in these processing plants could be clustered into two large groups. This clustering could be explained by the foods processed in each plant, since in P1 and P2 (group 2), ready-to-eat meal products (e.g., lasagnas and pizzas) were manufactured, while in P3 and P4 (group 1), both processed meat products (e.g., sausages, ham, and mortadella). Furthermore, processing plants included in the same group had raw materials in common (e.g., dairy products in P1 and P2; meat in P3 and P4), suggesting the possible influence of these materials on resident microbiota.

On the other hand, Pseudomonas spp. were a common bacteria identified across plants. This bacterium is important in the industrial environments due to the negative impact it may have on food quality, its ability to form biofilm, and to survive and grow under harmful conditions [2]. Some strains of Pseudomonas spp. have the ability to grow at low temperatures (4–12 °C) [37] and to tolerate some disinfectants, especially quaternary ammonium compounds [18], which are broadly used during the sanitation process in the food industry [38], and all plants involved in this study used this product during the disinfection process. Moreover, some studies have demonstrated that the association of Pseudomonas spp. with other bacteria, such as L. monocytogenes [18] and S. aureus [39], favors the development of more compact biofilm, which probably makes them more resistant to C&D process [40]. However, we did not find these bacteria cohabiting in the studied biofilms. On the other hand, we found cohabitation with lactic acid bacteria, which is an association that has been previously identified in biofilms [41].

In this study, Pseudomonas spp. were the dominant bacteria in biofilms in samples from P1 to P3 (more frequency and abundance), in comparison with other bacteria that cohabited these plants. Our results are in agreement with previous studies in which this bacterium was found to be the most predominant one in samples from food contact surfaces [42]. Some authors have reported the dominance of Pseudomonas spp. over other bacteria during biofilm formation in vitro. Langsrud et al., (2016) showed that Pseudomonas spp. always prevailed on the formed biofilms when was used a cocktail of different bacterial species, between 11 and 14 strains, isolated from food processing plants [9].

Acinetobacter spp. were another bacterium that predominated in the biofilms studied. It was present in most of the points sampled across the four food processing plants. This genus is often reported as a persistent bacterium after C&D process, not only to produce biofilm for its ability but also to survive under industrial environmental conditions [43]. Acinetobacter spp. may form biofilm under dynamic conditions, which gives a growth advantage over other bacteria. A study showed that E. coli O157:H7 strain, a bacterium without the ability to form biofilm under dynamic conditions, increased its count up to 400 times after it was embedded in the biofilm formed by A. calcoaceticus [19]. Our results showed that Acinetobacter spp. were the dominant genus in biofilms from P4, both in frequency and abundance. Nonetheless, according to previous studies, this behavior could be transient, since during an in vitro experiment, the authors observed that the predominance changed as the biofilm matured, initially identifying the predominance of Acinetobacter spp. (first 4 days) and of Pseudomonas putida at the end of the study [22]. The competition for limiting nutrient sources could be the cause of transitional microbial compositions of biofilms. Previous studies have demonstrated that both Acinetobacter spp. and Pseudomonas spp. can often be found in the same environment because these bacteria have similar characteristics, such as biofilm forming ability and antimicrobial resistance [44]. However, Acinetobacter spp. are less frequent in the production environment and have less impact on product quality [33].

Unlike P1, P3, and P4, where Gram-negative bacteria were the predominant ones, in P2, the Gram-positive bacterium, Lactobacillus spp., was the most frequent, being the only LAB with this abundance. This fact was enough to generate a change at the phylum level, since P2 was the only plant where the Firmicutes presented a very high proportion. LAB is a group of microorganisms capable of growing at low temperatures, and some species are relatively tolerant to stress conditions [45], thus providing them with a high food spoilage ability [46, 47]. Sample points where Lactobacillus genus predominated neither Pseudomonas spp. nor Acinetobacter spp. were present, at least among the top five more abundant bacteria. Therefore, Gram-positive bacteria prevailed in most sample points from P2 (72%), suggesting dominance and persistence of these bacteria over Gram-negative bacteria. Evidence has shown that LAB are predominant in dairy industries, where they are used as a starter culture [33]. This fact could explain the results obtained in the P2 because in that plant, dairy products (cheese and milk) were used as raw material. However, this was not observed in P1 samples, despite also having used the same dairy products.

Enterobacteriaceae is a bacterial family of great importance in the food industry because they are indicators of inappropriate hygiene practices. Bacteria such as Salmonella spp. and E. coli O157:H7 are considered foodborne pathogens, and some species are recognized as food spoilers. [48,49,50]. The main genera of the Enterobacteriaceae family found in this study were Escherichia spp. and Serratia spp. In agreement with our results, some authors have also reported on Escherichia genus in food processing environments [51]. High prevalence of Serratia spp. agrees with that reported in other studies, where this genus was more identified in different food processing plants [37]. Interestingly, a study in a milk processing plant showed that Serratia spp. presented greater ability of adherence and biofilm formation than Pseudomonas spp. [52, 53]. In addition, it has been evidenced that this bacterium may present high resistance to disinfectants and ability to grow at low temperatures, which allows it to compete with other psychrotrophic bacteria [53].

The resident microbiota in a specific environment has been also related with type of food, its composition, and nutrient availability for bacteria [54]. For instance, a study carried out in a dairy plant environment reported a core microbiota shaped by LAB, where Lactobacillus spp. showed a negative correlation with Pseudomonas spp. and Acinetobacter spp. According to the authors, these results could indicate that persistence of LAB in the environment helps to control the food spoilers [55]. In this study, similar results were obtained from P1, since LABs, specifically the Lactobacillus genus, were the predominant bacteria (57.1%) in most biofilms evaluated. In contrast to dairy plant environments, a study reported that bacteria, such as Psychrobacter, Rhodococcus, Leuconostoc, and Yersinia genera, had been detected in meat processing plant surfaces, with Yersinia spp. being the most abundant [45]. Unlike these results, we found Pseudomonas and Acinetobacter as the predominant genera across meat processing plants, P3 and P4. This is in agreement with other studies of bacterial spoilers in meat plant environments, where these bacteria predominated on the evaluated surface [12, 56].

Conclusions

The biofilms characterized from samples from the four processing plants showed Pseudomonas spp. and Acinetobacter spp. as the dominant bacteria, which are broadly recognized as strong biofilm formers. However, these bacteria cohabited with spoiler microorganisms of interest to the food industry such as Leuconostoc spp., Lactobacillus spp., Streptococcus spp., and Enterobacteriaceae.

Although in our study, foodborne pathogens were not detected in the biofilms, we cannot ignore the protector and synergic effect that offer the biofilms for these microorganisms, which represents a constant risk for the safety and quality of processed foods.

The mapping of the bacteria that inhabit on surface of the food producing areas is necessary to estimate the microbiological risk of each environment and to optimize the C&D protocols.